The continuum of Drosophila embryonic development at single-cell resolution

INTRODUCTION: Single-cell technologies are a powerful means of studying metazoan development, enabling comprehensive surveys of cellular diversity at profiled time points and shedding light on the dynamics of regulatory element activity and gene expression changes during the in vivo emergence of each cell type. However, nearly all such whole-embryo atlases of embryogenesis remain limited by sampling density—i.e., the number of discrete time points at which individual embryos are harvested and cells or nuclei are collected. Given the rapidity with which molecular and cellular programs unfold, this limits the resolution at which regulatory transitions can be characterized. For example, in the mouse, there are typically 6 to 24 hours between sampled embryonic time points—gaps within which massive molecular and morphological changes take place. RATIONALE: To construct an ungapped representation of embryogenesis in vivo, we would ideally sample embryos continuously. Although this is not practical for most model organisms, it is potentially possible in Drosophila melanogaster, where collections of timed and yet somewhat asynchronous embryos are easy to obtain, such that, at least in principle, one can achieve arbitrarily high temporal resolution. Drosophila could therefore serve as a test case to develop a framework for the inference of continuous regulatory and cellular trajectories of in vivo embryogenesis. Because Drosophila is a preeminent model organism that has yielded many advances in the biological and biomedical sciences, obtaining a single-cell atlas of Drosophila embryogenesis is also an important goal in itself. This includes its embryonic development, where the use of this model in conjunction with powerful genetic tools has transformed our understanding of the mechanisms by which developmental complexity is achieved, in addition to uncovering many general principles of both genetic and epigenetic gene regulation. RESULTS: We profiled chromatin accessibility in almost 1 million nuclei and gene expression in half a million nuclei from eleven overlapping windows spanning the entirety of embryogenesis (0 to 20 hours). To exploit the developmental asynchronicity of embryos from each collection window, we applied deep neural network-based predictive modeling to more-precisely predict the developmental age of each nucleus within the dataset, resulting in continuous, multimodal views of molecular and cellular transitions in absolute time. With these data, the dynamics of enhancer usage and gene expression can be explored within and across lineages at the scale of minutes, including for precise transitions like zygotic genome activation. CONCLUSION: This Drosophila embryonic atlas broadly informs the orchestration of cellular states during the most dynamic stages in the life cycle of metazoan organisms. The inclusion of predicted nuclear ages will facilitate the exploration of the precise time points at which genes become active in distinct tissues as well as how chromatin is remodeled across time.